Skip to content

Mingbird v2.1.1

Choose a tag to compare

View all tags
@Gustor-Wang Gustor-Wang released this 07 Oct 12:55
· 13 commits to main since this release
v2.1.1
2a13e59

鸣鸟 Mingbird v2.1.1

🇨🇳 中文在上 / English below.

首个多引擎版本。测试 540 项全绿(514 + 26 新增)。

新增

  • 本地引擎不再只有 Ollama——任何 OpenAI 兼容本地服务都能当后端:设置(⚙)里选 LM Studio(:1234/v1)、llama.app(llama.cpp 官方桌面应用,:9931/v1)、llama-server(:8080/v1)或自定义端点,API Key 可空。OpenAI 路径为真流式(SSE);服务端没配 reasoning 解析器时也能分出思考通道(行内 <think> 标签实时剥离,跨流式分片撕裂也不误判);流式工具调用按 index 增量拼装。模型下拉、状态灯、自愈刷新与 WebUI 全部随引擎自动切换。已经在用 LM Studio 或 llama.app?可以不装 Ollama,设置里选引擎即可。
  • 语法宽容扩围——文本形态工具调用外面包着的 ```json/```tool 围栏与 <tool_call> 标签壳,现在先剥离再交给既有挽救链解析(裸 JSON、函数表达式、tokenizer 占位符引号此前已覆盖)。
  • 输出上限新增 16K、32K 与不限——Ollama 引擎下"不限"映射为 num_predict: -1;OpenAI 兼容端点整体省略 max_tokens 字段(用服务端默认)。

Ollama 仍是默认引擎并保留全部专属能力(自动拉起、keep-alive、思考三态);连续 500 自动重启仍仅限 Ollama。

English

First multi-engine release. Tests: 540 green (514 + 26 new).

New

  • Local engines beyond Ollama — any OpenAI-compatible local server can now be the backend: pick LM Studio (:1234/v1), llama.app (the official llama.cpp desktop app, :9931/v1), llama-server (:8080/v1) or a custom endpoint in Settings (⚙), API key optional. The OpenAI path does true SSE streaming, separates the thinking channel even when the server runs no reasoning parser (inline <think> tags are stripped live, torn across stream chunks included), and assembles streamed tool calls incrementally. Model listing, the status light, self-healing and the WebUI follow the engine automatically. Already running LM Studio or llama.app? Skip Ollama — pick your engine in Settings.
  • Syntax leniency widened — fenced ```json/```tool blocks and <tool_call> tag shells around text-form tool calls are now stripped before the existing salvage chain parses them (bare JSON, function expressions and tokenizer placeholder quotes were already covered).
  • Output limit: 16K, 32K and Unlimited — on Ollama, Unlimited maps to num_predict: -1; on OpenAI-compatible endpoints the max_tokens field is omitted entirely (server default applies).

Ollama stays the default engine and keeps its exclusive extras (auto-start, keep-alive, thinking tri-state); the consecutive-500 auto-restart remains Ollama-only.

Install / 安装

Windows：下载 Mingbird-v2.1.1-CN-Setup.exe（中文）或 Mingbird-v2.1.1-EN-Setup.exe（English），双击安装。需要 Ollama 或任何 OpenAI 兼容本地服务；2~9B 小模型核显即可跑，全程数据不出本机。
Windows (EN): download Mingbird-v2.1.1-CN-Setup.exe (Chinese UI) or Mingbird-v2.1.1-EN-Setup.exe (English UI) and run the installer. Requires Ollama or any OpenAI-compatible local server; 2–9B models run on integrated graphics, and your data never leaves the machine.

macOS / Linux（实验性）：下载 mingbird-v2.1.1-macos-arm64.tar.gz / mingbird-v2.1.1-macos-x64.tar.gz / mingbird-v2.1.1-linux-x64.tar.gz，解压后运行 LocalAgent。系统语言自动识别（中文系统直接显示中文；英文系统输出自动跟随英文。也可在可执行文件旁建 app_lang.txt 写 zh/en 强制）。
macOS / Linux (experimental, EN): download the tar.gz for your arch, extract, run LocalAgent. System language is auto-detected (English systems get English output); you can also force it by placing an app_lang.txt (zh/en) next to the executable.

From source：pip install -r requirements.txt → python agent_gui.py（Python 3.12 推荐）。 / From source (EN): pip install -r requirements.txt → python agent_gui.py (Python 3.12 recommended).

Assets 7
Loading